main.py
No OneTemporary
Actions

Subscribers

None

File Metadata

Created: Fri, Jan 3, 06:57

main.py
View Options

	import warnings
	from sklearn.model_selection import GridSearchCV
	from sklearn.preprocessing import LabelEncoder
	from sklearn.metrics import classification_report
	from Z_HELPERS.DM_communication import *
	from data_processing import *
	from feature_extraction import *
	from model_training import *
	from visualization import *
	from optuna_objectives import *
	import pickle
	# Ignore future warnings
	warnings.filterwarnings("ignore", category=FutureWarning)

	def baseline_model_model1(normalized_transmission, interval_min, interval_max):
	# Baseline model will be our first implementation and will not use any learning method
	# Based on assumptions about gram classification in function of the mean value of normalized frequency
	mean_value = np.mean(normalized_transmission)
	# Classify based on the mean value
	if interval_min >= mean_value:
	return 1 # Gram-positive
	elif mean_value >= interval_max:
	return 0 # Gram-negative
	else:
	if interval_max - mean_value >= mean_value - interval_min:
	return 1
	else:
	return 0

	# Predict the nearest cluster for bacteria family classification
	def predict_bacteria_family_model1(normalized_transmission, cluster_centers):
	mean_transmission = np.mean(normalized_transmission)
	std_transmission = np.std(normalized_transmission)
	min_distance = float('inf')
	predicted_bacteria = None
	for bacteria, center in cluster_centers.items():
	distance = np.sqrt((mean_transmission - center[0]) 2 + (std_transmission - center[1]) 2)
	if distance < min_distance:
	min_distance = distance
	predicted_bacteria = bacteria
	return predicted_bacteria


	if __name__ == '__main__':
	print("Loading data from pickle file...")
	with open('data.pkl', 'rb') as f:
	docs = pickle.load(f)
	print("Data loaded successfully.")
	modelnb= input("Type the number of the model you want to use : ")
	# List of problematic values to exclude
	problematic_names = [
	'ns139_p8_2', 'ys134_p7_1', 'pp46_p3_0', 'li142_p2_3',
	'li142_p2_2', 'li142_p3_4', 'li142_p3_5', 'se9_p2_0', 'se9_p7_1'
	]
	display_names = {
	'bs134': 'B. subtilis',
	'coli': 'E. coli',
	'li142': 'L. innocua',
	'ns139': 'N. sicca',
	'pp46': 'P. putida',
	'pp6': 'Pseudomonas putida',
	'se26': 'S. epidermidis B',
	'se9': 'S. epidermidis A',
	'ys134': 'Y. ruckeri'
	}


	#model 1 not using any learning
	# Model 1 not using any learning
	if modelnb == '1':
	normalized_docs = normalize_data_model1(docs)
	filtered_docs = [doc for doc in normalized_docs if
	doc['name'] not in problematic_names and doc['bacteria'] != 'pp6']
	train_docs, test_docs = split_data(filtered_docs)
	interval_min, interval_max = calculate_means_and_interval_model1(train_docs)
	print(f"Interval: ({interval_min}, {interval_max})")

	# Gram type classification
	true_labels = []
	predicted_labels = []
	for doc in test_docs:
	normalized_transmission = doc['transmission_normalized']
	true_label = doc['label']
	predicted_label = baseline_model_model1(normalized_transmission, interval_min, interval_max)
	true_labels.append(true_label)
	predicted_labels.append(predicted_label)

	accuracy = np.mean(np.array(true_labels) == np.array(predicted_labels))
	print(f'Baseline model accuracy (Gram type): {accuracy * 100:.2f}%')

	# Bacteria family classification
	cluster_centers = find_cluster_centers(train_docs)
	#plot_training_clusters(train_docs, cluster_centers, display_names)
	true_bacteria_labels = []
	predicted_bacteria_labels = []
	for doc in test_docs:
	normalized_transmission = doc['transmission_normalized']
	true_bacteria = doc['bacteria']
	predicted_bacteria = predict_bacteria_family_model1(normalized_transmission, cluster_centers)
	true_bacteria_labels.append(true_bacteria)
	predicted_bacteria_labels.append(predicted_bacteria)

	bacteria_accuracy = np.mean(np.array(true_bacteria_labels) == np.array(predicted_bacteria_labels))
	print(f'Baseline model accuracy (Bacteria type): {bacteria_accuracy * 100:.2f}%')
	plot_all_clusters(train_docs, test_docs, cluster_centers, display_names, true_bacteria_labels,predicted_bacteria_labels)


	# Scatter plot for Gram type classification



	# Call the function to plot Gram type classification
	plot_gram_type_classification(train_docs, interval_min, interval_max)
	from sklearn.metrics import confusion_matrix, ConfusionMatrixDisplay


	# Confusion matrix for bacteria family classification with improved formatting



	# Call the function to plot the confusion matrix
	plot_confusion_matrix(true_bacteria_labels, predicted_bacteria_labels, display_names)


	# Function to plot cluster center calculation illustration



	# Call the function to plot cluster center calculation illustration
	plot_cluster_center_calculation(train_docs, cluster_centers, display_names)

	# Model 2 using basic, simple learning methods

	elif modelnb == '2':
	normalized_docs = normalize_data_model2(docs)
	filtered_docs = [doc for doc in normalized_docs if
	doc['name'] not in problematic_names and doc['bacteria'] != 'pp6']
	chunked_docs = chunk_time_series(filtered_docs, num_chunks=6)

	# Augment the data
	augmented_docs = augment_data(chunked_docs, noise_level=0.05, shift_max=10)

	# Perform repeated random splitting and evaluation
	gram_accuracies, bacteria_accuracies = repeated_random_splitting_evaluation(augmented_docs, iterations=2)

	# Print the accuracy scores for each iteration
	print("Gram type classifier cross-validation accuracies for each iteration:")
	print(gram_accuracies)
	print(f"Mean accuracy: {np.mean(gram_accuracies):.2f} ± {np.std(gram_accuracies):.2f}")

	print("Bacteria type classifier cross-validation accuracies for each iteration:")
	print(bacteria_accuracies)
	print(f"Mean accuracy: {np.mean(bacteria_accuracies):.2f} ± {np.std(bacteria_accuracies):.2f}")

	# Model 3: Using CatBoost with Optuna
	elif modelnb == '3':
	normalized_docs = normalize_data_model2(docs)
	filtered_docs = [doc for doc in normalized_docs if
	doc['name'] not in problematic_names and doc['bacteria'] != 'pp6']
	chunked_docs = chunk_time_series(filtered_docs, num_chunks=6)
	augmented_docs = augment_data(chunked_docs, noise_level=0.05, shift_max=10)

	train_docs, test_docs = split_data(augmented_docs)

	train_features, train_gram_labels, train_bacteria_labels = extract_features_model2(train_docs)
	test_features, test_gram_labels, test_bacteria_labels = extract_features_model2(test_docs)

	label_encoder = LabelEncoder()
	encoded_train_bacteria_labels = label_encoder.fit_transform(train_bacteria_labels)
	encoded_test_bacteria_labels = label_encoder.transform(test_bacteria_labels)

	X_train, X_test, y_train, y_test = train_features, test_features, train_gram_labels, test_gram_labels
	X_train_bacteria, X_test_bacteria, y_train_bacteria, y_test_bacteria = train_features, test_features, encoded_train_bacteria_labels, encoded_test_bacteria_labels

	objective = ObjectiveXGB(X_train, X_test, y_train, y_test, X_train_bacteria, X_test_bacteria, y_train_bacteria,
	y_test_bacteria)

	study_gram = optuna.create_study(direction='maximize')
	study_gram.optimize(objective.objective_xgb_gram, n_trials=50)

	print(f"Best trial (Gram type): {study_gram.best_trial.value}")
	print("Best hyperparameters (Gram type): ", study_gram.best_trial.params)

	best_params_gram = study_gram.best_trial.params
	xgb_gram = XGBClassifier(**best_params_gram, use_label_encoder=False, eval_metric='logloss')
	xgb_gram.fit(X_train, y_train)

	train_accuracy_gram = xgb_gram.score(X_train, y_train)
	print(f'Training accuracy for Gram type classifier: {train_accuracy_gram:.2f}')

	y_pred_gram = xgb_gram.predict(X_test)
	print("Gram Type Classification Report:")
	print(classification_report(y_test, y_pred_gram, zero_division=0))

	study_bacteria = optuna.create_study(direction='maximize')
	study_bacteria.optimize(objective.objective_xgb_bacteria, n_trials=50)

	print(f"Best trial (Bacteria type): {study_bacteria.best_trial.value}")
	print("Best hyperparameters (Bacteria type): ", study_bacteria.best_trial.params)

	best_params_bacteria = study_bacteria.best_trial.params
	xgb_bacteria = XGBClassifier(**best_params_bacteria, use_label_encoder=False, eval_metric='mlogloss',
	objective='multi:softmax', num_class=len(np.unique(y_train_bacteria)))
	xgb_bacteria.fit(X_train_bacteria, y_train_bacteria)

	train_accuracy_bacteria = xgb_bacteria.score(X_train_bacteria, y_train_bacteria)
	print(f'Training accuracy for Bacteria type classifier: {train_accuracy_bacteria:.2f}')

	y_pred_bacteria = xgb_bacteria.predict(X_test_bacteria)
	print("Bacteria Type Classification Report:")
	print(classification_report(y_test_bacteria, y_pred_bacteria, target_names=label_encoder.classes_,
	zero_division=0))

	gram_cv_scores = cross_val_score(xgb_gram, X_train, y_train, cv=5, scoring='accuracy')
	print(
	f'Gram type classifier cross-validation accuracy: {gram_cv_scores.mean():.2f} ± {gram_cv_scores.std():.2f}')

	bacteria_cv_scores = cross_val_score(xgb_bacteria, X_train_bacteria, y_train_bacteria, cv=5, scoring='accuracy')
	print(
	f'Bacteria type classifier cross-validation accuracy: {bacteria_cv_scores.mean():.2f} ± {bacteria_cv_scores.std():.2f}')

	test_accuracy_gram = xgb_gram.score(X_test, y_test)
	test_accuracy_bacteria = xgb_bacteria.score(X_test_bacteria, y_test_bacteria)
	print(f'Test accuracy for Gram type classifier: {test_accuracy_gram:.2f}')
	print(f'Test accuracy for Bacteria type classifier: {test_accuracy_bacteria:.2f}')

	if train_accuracy_gram > test_accuracy_gram:
	print("Potential overfitting detected in Gram type classifier.")
	elif train_accuracy_gram < test_accuracy_gram:
	print("Potential underfitting detected in Gram type classifier.")

	if train_accuracy_bacteria > test_accuracy_bacteria:
	print("Potential overfitting detected in Bacteria type classifier.")
	elif train_accuracy_bacteria < test_accuracy_bacteria:
	print("Potential underfitting detected in Bacteria type classifier.")

main.pyNo OneTemporaryActions

File Metadata

main.pyView Options

Event Timeline

main.py
No OneTemporary
Actions

main.py
View Options